Using ARCAP time-frequency representations for decision

We address the problem of estimating and classifying multicomponent narrow band signals using ARCAP Time-Frequency Representations (TFRs). ARCAP TFRs are build using firstly, an instantaneous frequency estimator and, secondly, an instantaneous power estimator, and results in a set of points without any time links. The spectral trajectories have then to be extracted, and we apply a Kalman filter. The trajectories parameters (chirps slopes, beginning and ending instants and frequencies…) enable the implementation of a supervised classification procedure. Results are given for mono and multicomponent chirp signals.